1

Casting Light on Energy Efficiency - Evidence on Consumer Inattention

and Imperfect Information

Matthias Rodemeier*1, Andreas Löschel*, Roland Kube*

* University of Münster, Chair of Microeconomics, esp. Energy and Resource

Economics, Am Stadtgraben 9, D-48143 Münster, Germany

January 2017

Abstract

We investigate consumer inattention and imperfect information regarding the financial benefits of

energy-efficient lighting using a randomized controlled trial with 1,084 observations. Results suggest

that subjects generally know about cost savings of LED bulbs - the central lighting technology of the

future - but largely underestimate the magnitude of these savings. As a result, stated willingness-to-pay

for an LED bulb increases on average by 2.53€ through the provision of information on expected lifetime

costs. Consumers also confound technology attributes of energy-efficient alternatives, which further

explains low adoption rates of the LED technology.

Highlights

We investigate informational and attentional biases in purchase decisions about an innovative

lighting technology using a randomized controlled trial with hypothetical choices for a large

sample in Germany.

We find that stated willingness-to-pay for an LED bulb can on average be increased by 2.53€

through the provision of information on expected lifetime costs.

Consumers are confused about differences between energy-efficient alternatives and falsely

assign a negative attribute to LEDs.

Keywords

Imperfect Information, Inattention, Energy Efficiency Gap, Experimental Economics

JEL Codes

D03, D12, D83, Q41, Q48

1. Introduction

Residential lighting is one of the largest electricity end-users in European households and still subject

to immense savings potentials, especially when light-emitting diode (LED) bulbs are taken for

replacement (De Almeida et al., 2011). Household lighting is also ranked among the most cost-efficient

means to reduce externalities from CO2 emissions (IPCC, 2007). Yet, the adoption of efficient lighting

by consumers remains slow, which is particularly puzzling as LED bulbs provide large financial benefits

relative to classical alternatives.

Theoretical explanations of this phenome include (rational) inattention to energy efficiency, imperfect

information, high discount rates or simply strong preferences for other product attributes.2 This paper

tests for these different causes by using a randomized controlled trial with an information treatment

based on Allcott & Taubinsky (2015) for the US market. In their study, the authors find that consumers

1 Corresponding author at: Chair of Microeconomics, esp. Energy and Resource Economics. Am

Stadtgraben 9, 48143 Münster, Germany. Tel.: +49 251 8322978. Email address: mattthias.rodemeier@wiwi.uni-muenster.de 2 For an overview on potential causes of a so-called “Energy Efficiency Gap” see Gerarden et al.

(2015).

2

undervalue energy-efficient compact fluorescent bulbs (CFL) due to a lack of energy literacy and

possibly inattention. To our knowledge, we provide the first related evidence for an even more relevant

lighting technology with a focus on the German market.

LEDs are three times more energy-efficient than CFLs and promise even higher cost advantages in the

future due to constantly decreasing prices (McKinsey & Company, 2012). In addition, LEDs constitute a

closer substitute to traditional incandescent bulbs than CFLs as they include no (potentially health-

damaging) mercury content and as they reach full brightness immediately. This differentiation is

important as consumers should be less inattentive to differences in energy efficiency when product

attributes in other dimensions are similar (Sallee, 2014).

We test for undervaluation of LED bulbs resulting from consumer biases in a randomized controlled trial

with hypothetical consumption choices. The analyzed data constitutes a notably large subsample

(N=1,084) of a country which is not only the largest economy in the European Union, but also seen as

a leader in current energy transformation policies (IRENA, 2015).

2. Experimental Design

Between June and July 2016 people were invited to participate in an online questionnaire via email

distributors of German universities and through announcements on social networks. Our sample is

consequently drawn from a young and rather well-educated subpopulation. Participation was

incentivized through a lottery of cash prizes and vouchers for an online shop.

Upon opening the online questionnaire, subjects were randomly assigned to treatment and control

group. The survey started with a short introductory screen (see B.1) and a subsequent screen showing

different lamp types in a modern living room (B.2 and B.3). The latter was designed to raise subjects’

interest for the survey, which is generally known to increase the reliability of survey responses (Warwick

& Lininger, 1975). Participants were then asked to imagine they needed a new light bulb and make

hypothetical purchase choices between a 40W incandescent and a 5W LED at varying prices (B.6 and

B.7). As depicted in B.7, subjects had to fill in a multiple price list in which the price of the LED increased

in ascending order from 0.30€ to 20.30€ while the price of the incandescent was fixed to 1.30€. We

define the subjects’ relative Willigness-to-pay (WTP) for the LED as the average between the two LED

prices at which the subject switches from choosing the LED to choosing the incandescent, minus the

price of the incandescent bulb.3

For individuals in the treatment group, an additional screen prior to the purchase decision appeared (B.4

and B.5) and offered written and graphical information about average differences in electricity and

replacement costs between the two bulbs. Following Allcott & Taubinsky (2015), we assume that this

intervention eliminates any distortion in consumer choices resulting from inattention to or biased beliefs

about the energy efficiency of the two bulbs. Since the only difference between treatment and control

group is this information screen, systematic differences in WTP indicate undervaluation of the financial

benefits from energy efficiency.

Given that WTP is determined using stated preferences, our estimates are vulnerable to hypothetical

bias. Note, however, that estimates from stated preferences are found to be significantly less biased for

private goods than for public goods as consumers are more familiar with such products on markets (List

& Gallet, 2001).

The survey involved further questions on socioeconomic variables, implicit discount rates, other

preferences for light bulbs and psychological characteristics (see B.10 to B.19).

3 For instance, if the consumer purchased the LED at 3.30€ but switches to the incandescent as soon as the LED costs 4.30€, we define her WTP for the LED as (3.30€ + 4.30€)/2 − 1.30€ = 2.50€.

3

3. Results

The dataset contains 1,084 observations and mostly consists of students (87%). Table A.1 shows that

treatment and control group are well balanced in individual covariates and confirms successful

randomization.

Table 1 presents basic OLS estimates. We include all subjects with non-censored WTP, meaning that

they implicitly revealed their WTP by switching between the incandescent and the LED at some price in

the presented price list.4 The average treatment effect is a statistically significant increase in WTP for

the LED bulb of 2.71€. This effect decreases only slightly to 2.53€ when controlling for observable

characteristics (Column 2). Our estimates are fairly similar to the incentive-compatible estimate by Allcott

& Taubinsky (2015) who find an increase in WTP for CFLs of $2.54 (≈2.02€ at the time of the survey)

for the US sample. A larger treatment effect is plausible in our case because LEDs save substantially

more energy costs than CFLs and choices in our study were of hypothetical nature.

Table 1

OLS Estimates of Treatment Effect

Dependent variable: Relative willingness-to-pay for the LED bulb

(1) (2)

Treatment 2.705 2.532 (0.248)*** (0.256)*** Observables No Yes Constant 3.735 2.562 (0.127)*** (1.194)** R2 0.13 0.19 N 932 932

Notes: Robust standard errors are in parentheses. Significance levels are given by * p<0.1; ** p<0.05; *** p<0.01.

Figure 1 provides a comparison of demand curves for LED bulbs between control and treatment group.

At the typical relative market price of these two bulbs (approximately 6€ in Germany) the share of

consumers choosing the LED more than doubles from 19 to 45 percent as a result of the information

treatment.

In order to identify whether our treatment effect results from increased information about or just

inattention to energy efficiency, we ask all subjects additional questions on energy literacy. Subjects

were asked which of the two bulbs had lower operating costs (B.10) and how much lower these costs

were for 15 years of usage (B.11 and B.12). The results in column (1) and (2) of Table A.2 are obtained

by using probit regressions to regress the binary variables “Belief: LED is cheaper” and “Belief: LED

saves 120€” on the treatment. The first dependent variable is equal to 1 if the subject correctly answered

the LED was cheaper than the incandescent, and zero otherwise. Analogously, the second variable

takes on the value 1 if the subject answered “120€” on the question regarding how much the LED saves

compared to the incandescent, and zero if she chose any other answer. Being part of the treatment

4 Of the entire sample, 152 subjects preferred the same bulb at any given price and had to type in its

minimum/ maximum WTP for the LED in an additional field (B.8). Given that these specific subjects were able to state an arbitrarily large WTP, we analyze this subsample carefully. If we include these subjects, the average treatment effect increases to 7.82€. However, the median treatment effect only increases to 3.27€, indicating that this drastic increase in the average treatment effect is driven by a few subjects who reported an exceptionally large WTP.

4

group increased the probability of answering that LEDs are less expensive in usage by 3.1 percentage

points. While this effect is highly significant from a statistical perspective, its economic magnitude is

relatively small. Even in the control group, 95 percent of subjects answered that using the LED was

cheaper. Much larger differences exist when it comes to the accuracy of savings beliefs. Column (2)

implies that the treatment increased the probability of giving the “correct” answer on expected cost

differences by 25.2 percentage points for the treatment group. Only 16 percent in the control group had

savings beliefs which were equal to the estimated average savings of around +120€. Consumers appear

to know LEDs have lower usage costs in general, but have biased beliefs about the magnitude of the

financial savings.

Figure 2 illustrates the density functions of savings beliefs between treated and non-treated subjects.

The density function of the control group is centered around values closer to zero and involves a notably

larger variance.

Figure 1

Empirical Demand Curves

Notes: The relative price is defined as the price of the LED minus the price of the incandescent.

In addition, we asked subjects about the importance of factors that have influenced their hypothetical

purchase decision using a Likert-Scale (B.14). Results are used as regressors for WTP in Table A.3.

Consumers who put a high emphasis on the bulb’s CO2 emissions, its energy consumption and its

lifetime have a significantly higher WTP for the LED, unlike consumers who focus on the initial purchase

price. Interestingly, consumers who placed high importance on the time until the bulb reaches full

brightness also show a significantly lower WTP for the LED. Note, however, that both incandescents

and LEDs immediately reach full brightness. A long warm-up time is characteristic for CFLs and found

to be an unpopular feature among consumers in other studies (Rasmussen et al., 2007; Wall & Crosbie,

2009). Since LEDs are relatively new on the lighting market, this may suggest that consumers confound

LEDs with CFLs or assume energy-efficient bulbs to need more time to warm up in general. The finding

that consumers appear to have biased beliefs about differences between energy-efficient technologies

is a non-negligible result since it could translate into other markets for energy-using durables.

Another hypothesis to be tested is that consumers who discount future utility at larger rates should be less inclined to purchase the LED, as energy savings are benefits accruing in the future. We address this conjecture by asking consumers whether they hypothetically prefer receiving 100€ today or varying amounts between 100€ and 200€ in one year (see B.13). The discount rate is defined as 𝑖 =

(𝑠𝑤𝑖𝑡𝑐ℎ𝑖𝑛𝑔 𝑝𝑜𝑖𝑛𝑡

100€⁄ ) − 1, where the switching point is the average of the two monetary amounts in

one year at which the consumer switches from preferring money today to money in the future. Column

5

1 in Table A.4 regresses WTP for the LED on the implicit discount rate and finds that an increase in the discount rate by 10 percentage points is associated with a statistically significant decrease in average WTP of 0.11€. The average discount rate of the analyzed sample is 23%. Economic intuition is supported by columns (2)-(4), where we find evidence that purchase decisions of subjects with higher discount rates are less influenced by the bulbs’ energy costs, its lifetime and its final disposal.

Figure 2

Density Functions of Savings Beliefs

Notes: The figure illustrates the Epanechnikov kernel density functions of savings beliefs elicited by questions B.11 and B.12.

4. Conclusion

Our work provides evidence for significant undervaluation of LED bulbs in Germany resulting from biased beliefs about financial benefits of energy efficiency. Given that we have analyzed a subsample with an above-average educational level, these effects are likely to be even larger for the entire population. Additional results suggest that consumers with higher discount rates are more likely to favor incandescents and that the adoption of LEDs may further be hampered as consumers are confused about differences between energy-efficient alternatives.

Our results are also relevant from a political perspective since the European Union considers LEDs as the most important alternative to traditional incandescents and established the “European LED Quality Charter” to improve consumer acceptance of LED bulbs (European Commission, 2012). The presented findings provide ground for a discussion on information policies as adequate means to promote the adoption of energy-efficient lighting and its associated benefits regarding externality reductions.

Acknowledgments

We would like to thank the German online retailer Lampenwelt for financing the lottery among participants. We also thank Jörg Lingens and Madeline Werthschulte for highly valuable contributions, and Martin Skala for his organizational support.

6

References

Allcott, H. & Taubinsky, D. (2015). Evaluating Behaviorally-Motivated Policy: Experimental Evidence from the Light bulb Market. American Economic Review, 105(8), 2501-2538.

European Commission (2012). Energy Efficiency Status Report. Joint Research Centre. Institute for Energy and Transport.

De Almeida, A., Fonseca, P., Schlomann, B., & Feilberg, N. (2011). Characterization of the household electricity consumption in the EU, potential energy savings and specific policy recommendations. Energy and Buildings, 43 (8), 1884-1894.

Gerarden, T., Newell, R. G., & Stavins, R. N. (2015). Deconstructing the energy-efficiency gap: conceptual frameworks and evidence. American Economic Review, 105(5), 183-186.

IPCC (2007). Climate Change 2007: Mitigation. Contribution of Working Group III to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Metz, B., Davidson, O.R., Bosch, P.R., Dave, R., Meyer, L.A. (eds)], Cambridge University Press, Cambridge.

IRENA (2015). Renewable Energy Prospects: Germany. Remap 2030 analysis. International Renewable Energy Agency (IRENA), Abu Dhabi.

List, J. A., & Gallet, C. A. (2001). What experimental protocol influence disparities between actual and hypothetical stated values?.Environmental and Resource Economics, 20(3), 241-254.

McKinsey & Company (2012). Lighting the Way - Perspectives on the global lighting market. Second Edition. Vienna.

Rasmussen, T., Canseco, J., Rubin, R., & Teja, A. (2007). Are we done yet? An assessment of the remaining barriers to increasing compact fluorescent lamp installations and recommended program strategies for reducing them. In: Proceedings of 2007. European Council for an Energy Efficient Economy, 1951–1958. Online access: www.eceee.org/conference_proceedings/eceee/2007/Panel_9/9.301

Sallee, J. M. (2014). Rational inattention and energy efficiency. The Journal of Law and Economics, 57(3), 781-820.

Wall, R., & Crosbie, T. (2009). Potential for reducing electricity demand for lighting in households: An exploratory socio-technical study. Energy Policy, 37(3), 1021-1031.

Warwick, D. P., & Lininger, C. A. (1975). The Sample Survey: Theory and Practice. McGraw-Hill. New York.

7

Appendix A: Additional Tables

Table A.1 Balance of Observables between Treatment and Control Group

Treatment Group Mean

Control Group Mean

Difference (Treatment –

Control) (1) (2) (3)

Conservative 0.119 0.0962 0.0230 (0.324) (0.295) (0.0188) Social democrat 0.223 0.222 0.000628 (0.416) (0.416) (0.0253) Liberal 0.119 0.131 -0.0120 (0.324) (0.338) (0.0202) Leftist 0.121 0.145 -0.0240 (0.327) (0.353) (0.0207) Right-wing 0.00586 0 0.00586 (0.0764) (0) (0.00319) Ecological 0.172 0.156 0.0163 (0.378) (0.363) (0.0225) Other political affiliation

0.0195 (0.139)

0.0157 (0.125)

0.00380 (0.00799)

Not interested in politics

0.0977 (0.297)

0.107 (0.309)

-0.00899 (0.0185)

Statement on political affiliation denied

0.123 (0.329)

0.128 (0.334)

-0.00458 (0.0202)

Tenant 0.889 0.890 -0.00119 (0.315) (0.313) (0.0191) Homeowner 0.0938 0.0822 0.0116 (0.292) (0.275) (0.0172) Homeowner and tenant

0.0176 (0.132)

0.0280 (0.165)

-0.0104 (0.00914)

Customer of “green electricity”

0.270 (0.444)

0.299 (0.458)

-0.0294 (0.0275)

German basic school diploma (“Hauptschule”)

0 (0)

0.00175 (0.0418)

-0.00175 (0.00185)

German middle school diploma

0.00391 (0.0624)

0.00524 (0.0723)

-0.00134 (0.00413)

8

(“Realschule”) German high school diploma (“Abitur”)

0.637 (0.481)

0.626 (0.484)

0.0108 (0.0294)

Apprenticeship 0.0313

(0.174) 0.0385 (0.192)

-0.00721 (0.0112)

University degree

0.322 (0.468)

0.322 (0.468)

0.000587 (0.0285)

Statement on education denied

0.00586 (0.0764)

0.00699 (0.0834)

-0.00113 (0.00488)

Don’t know education degree

0 (0)

0 (0)

0 (0)

Female 0.576 0.570 0.00624 (0.495) (0.496) (0.0301) Male 0.424 0.430 -0.00624 (0.495) (0.496) (0.0301) Searching for employment

0.00195 (0.0442)

0.00175 (0.0418)

0.000205 (0.00261)

Employed 0.127 0.105 0.0221 (0.333) (0.307) (0.0194) Pupil 0.00195 0.00699 -0.00504 (0.0442) (0.0834) (0.00412) Student 0.869 0.879 -0.0102 (0.338) (0.326) (0.0202) Occupation not specified

0 (0)

0.00699 (0.0834)

-0.00699 (0.00369)

Notes: Column (1) and (2) have standard deviation in parentheses. Column (3) has standard errors in parentheses. Significance levels are given by * p<0.1; ** p<0.05; *** p<0.01.

9

Table A.2 Effect of Treatment on Savings Beliefs

Belief: LED is cheaper (marginal

effects obtained from probit regression)

Belief: LED saves 120€ (marginal effects obtained

from probit regression)

Median Savings Beliefs (in EUR)

(1) (2) (3)

Treatment 0.031 0.252 50.000 (0.012)** (0.023)*** (4.243)***

Constant 70.000 (2.916)*** N 1,084 1,084 1,084

Notes: Results in column (1) and (2) are marginal effects obtained from probit regressions. Estimates from column (3) are obtained through quantile regressions. Standard errors are in parentheses. Significance levels are given by * p<0.1; ** p<0.05; *** p<0.01.

Table A.3 Association of Factor Importance and Willingness-to-pay

Dependent variable: Relative willingness-to-pay for the LED bulb

(1) (2) (3) (4) (5) (6) (7) (8) (9)

Design of the bulb

0.106 (0.084)

Brightness 0.322

(0.123)***

CO2-emissions

0.265 (0.095)***

Energy costs

0.673 (0.119)***

Purchase price

-0.859 (0.126)***

Lifetime 0.538

(0.121)***

Mercury content

0.212 (0.095)**

Disposal 0.094

(0.102)

Warm-up time

-0.217 (0.097)**

Constant 4.520 3.525 4.088 2.191 8.158 2.740 4.267 4.546 5.633 (0.284)*** (0.528)*** (0.326)*** (0.492)*** (0.514)*** (0.500)*** (0.329)*** (0.295)*** (0.360)*** R2 0.00 0.01 0.01 0.03 0.05 0.02 0.01 0.00 0.01 N 880 913 816 904 926 891 724 751 848

Notes: Results are obtained from OLS regressions. Robust standard errors are in parentheses. Significance levels are given by * p<0.1; ** p<0.05; *** p<0.01.

10

Table A.4 Association of Discount Rates, Willingness-to-pay and Factor Importance

Relative willingness-to-

pay for the LED bulb

Importance of Bulb’s Energy

costs

Importance of Bulb’s Lifetime

Importance of Bulb’s Disposal

(1) (2) (3) (4)

Implicit discount rate (= 𝑖 × 100)

-0.011 (0.006)*

-0.005 (0.002)***

-0.004 (0.002)**

-0.006 (0.002)**

Observables No No No No Constant 5.112 4.156 4.144 2.717 (0.204)*** (0.052)*** (0.052)*** (0.076)*** R2 0.00 0.01 0.01 0.01 N 879 853 839 707

Notes: Results are obtained from OLS regressions. The average discount rate of the total sample is 23%. Robust standard errors are in parentheses. Significance levels are given by * p<0.1; ** p<0.05; *** p<0.01.

Appendix B: Instructions

All instructions were translated from German to English.

Figure B.1: Introduction Screen

11

Figure B.2: Distraction Screen 1 (Top of Screen)

Figure B.3: Distraction Screen 1 (Bottom of Screen)

12

Figure B.4: Treatment Screen (Top of Screen)

13

Figure B.5: Treatment Screen (Bottom of Screen)

Figure B.6: Purchase Decision (Top of Screen)

14

Figure B.7: Purchase Decision (Bottom of Screen)

Figure B.8: Question on Maximum Willingness-to-pay if larger than 20.30€

15

Figure B.9: Distraction Screen 2

Figure B.10: First Question on Savings Beliefs

16

Figure B.11: Second Question on Savings Beliefs if Answer to First Question was “Cheaper”

Figure B.12: Second Question on Savings Beliefs if Answer to First Question was “More Expensive”

17

Figure B.13: Question to elicit Implicit Discount Rates

Figure B.14: Question on Importance of Factors influencing the Purchase Decision

Figure B.15: Question on Political Affiliation

18

Figure B.16: Elicitation of Psychological Characteristics

Figure B.17: Question on Home-ownership

Figure B.18: Invitation to participate in the Lottery after completing the Survey

Figure B.19: Final Screen after completing the survey